What I want to do: I am trying to do NDVI for crop scouting in AZ, and I am using a modified Mobi...
Public Lab is an open community which collaboratively develops accessible, open source, Do-It-Yourself technologies for investigating local environmental health and justice issues.
Public Lab Research note
As an open source community, we believe in open licensing of content so that other members of the community can leverage your work legally -- with attribution, of course. By joining the Public Lab site, you agree to release the content you post here under a Creative Commons Attribution Sharealike license, and the hardware designs you post under the CERN Open Hardware License 1.1 (full text). This has the added benefit that others must share their improvements in turn with you.
What I want to do: I am trying to do NDVI for crop scouting in AZ, and I am using a modified Mobius camera with red filter.
My attempt and results: I keep loosing the PL page that has a screen shot of the proper settings for white balance. I am using R=310, G=500, B=700. This is the raw image that my settings captured.
Questions and next steps: I am in need of some help to see if my images are coming out correctly or not.
Why I'm interested: I want to be able to use the Modified Mobius with the PL red filter and get great NDVI images that have value.
When the NIR cut filter is removed from a camera, all three color channels capture NIR light in addition to the color they originally captured (R, G, or B). So NIR is mixed with visible light in all channels in unknown proportions.
Adding certain colored filters can allow one of the color channels to capture only NIR light and another to capture mostly visible light.
First of all thanks for your prompt reply. I have few more things which i am confused about:
1. If i put a red filter, so my red channel would mostly contain red spectrum while the blue will mostly contain infrared spectrum, what about the green channel, will i get green visible light in green channel, this is because i plan to compute NDWI index, for water content detection?
2. What is the purpose/need for custom white balance?
3. For blue filters, does it only allow blue spectrum and near infrared only?
4. If there are any related literatures, journals published by the Public labs team, please share the link.
The red filters typically used do not transmit much green light, so the green channel captures mostly NIR. Some yellow filters pass red and green, so the green channel captures green light. But some green light will also end up in the blue channel so the NIR channel will be more contaminated than with a red filter. You can read more and buy a piece of Wratten 15 filter here.
NDVI and NDWI rely on the difference between the visible and NIR reflection from foliage, so visible and NIR light must be measured in the same way. Converted consumer cameras are much less sensitive to NIR light than to visible light, and the filters used might not pass exactly the same amount of visible and NIR. So using the brightness values in the visible and NIR channels to compute NDVI will not produce meaningful results. A clever but crude way to adjust the relative brightness of the visible and NIR channels is to apply a custom white balance to the camera that exaggerates the NIR channel. If brightness in the NIR channel is sufficiently increased, the NDVI values can be in the correct range.
It's a little harder to make a filter that passes blue and NIR but none of the green and red in between. So the typical blue filters used probably have more contamination of visible light in the NIR channel (red) compared to the typical red filters used. Depending on the filter, the green channel can capture green and NIR.
Thanks for sharing the publication. My main purpose is to test the potential of calculating NDVI using modified infrared mobius actioncam (factory modified). My objective includes the efficiently stitching the multiple images (~ 80 images) collected from DJI Phantom 3 Professional and running the algorithm for calculating NDVI. However, i have not quite focused on calibrating the NDVI images due to which my initial flight data didn't work good. So, i trying to figure out best way to calibrate the images for NDVI.
I have mobius actioncam modified infrared with 87 mm Standard A lens fitted with red filter by public labs. But due to the wide angle lens i have huge distortion and fish eye effect so i plan to buy 6 mm modified infrared security lens from public labs. Which filter would you suggest me to add to my upcoming 6 mm security lens, as for now i have been suggested by Frank(a member of public labs to add the blue filter). I wanted to know your views before purchasing it.
Is the red filter from public lab infragram project similar to Wratten 15 filter as you mentioned above?
Stitching Mobius photos can be hard. The longer lens will help. Correcting the distortion afterwards can help. A high shutter speed and vibration damping can help. I assume you can also program the DJI to stop and hover for each photo. If so it would be a mistake not to do that.
The red filter from Public Lab is Rosco #19 which is superficially similar to the Wratten 25.
The blue filter from Public Lab is Rosco #2007 which is superficially similar to the Wratten 47 or Schott BG3.
The Wratten 15 is yellow-orange.
Reply to this comment...
Great.... For the accurate dense reconstruction with my mobius setup with wide angle lens, i have figured out best way. There is no option to work out with the shutter speed so i have made a mission plan with a speed of 9 m/s and i take timelapse every 5 sec which would be sufficient for getting good overlap. For, vibration removal i have setup something like this: https://www.youtube.com/watch?v=PuNhiJAVum0
I will also be taking two flights one after another, where i will maintain exactly perpendicular flight direction with +- 20 m change in altitude between the flights. This would help me in getting accurate stiching of multiple images (~ 90 images). After some pre and post calibration i think i will get better results with wide angle A lens mobius.
I will test this configuration and share my results in public labs.
Thank you Chrish. You guys are wonderful and passionate.